Embeddings Dimensionality Guide

Reference for embedding models: dimensions, size, performance benchmarks. OpenAI ada-002, Cohere, sentence-transformers.

Filter by Provider

Sort By

Model	Provider	Dims	Max Tokens	MTEB Score	Price/1M tokens	Notes

Dimensions

Number of Vectors

Precision

Higher dimensions generally capture more semantic nuance but require more storage
For most use cases, 768-1536 dimensions provide good quality-to-cost ratio
Consider Matryoshka embeddings (text-embedding-3) to reduce dimensions with minimal quality loss
Open-source models can be self-hosted to eliminate per-token API costs
MTEB (Massive Text Embedding Benchmark) provides standardized comparison scores

More Developer Tools tools at toool.cc